How do AltHold and Loiter feel in 4.7 vs 4.6?

ArduCopter Copter 4.7
1

The most controversial issue in 4.7 involves appears to be the addition of jerk limiting to the attitude controller. For many pilots they notice no difference between 4.6 and 4.7 but we have received some complaints that:include:

  1. the vehicle is less responsive in Loiter mode
  2. the vehicle wobbles in roll and pitch in Loiter, Guided and other modes using the position controller

If you see one of these problems we recommend reducing ATC_INPUT_TC from the default of 0.15 to 0.1

On the positive side we think that the vehicle’s attitude will be a little smoother and spikes in motor outputs should be reduced

Any and all feedback good or bad is greatly appreciated!

6 Likes
2

I use 4.7.0.dev while ago and see no difference or other problems in ALtHold or other modes.

1 Like
3

@rmackay9 this is only for small vehicles, right?

Bigger vehicles do not report any issues, right?

4

Hi @poruchik111, thanks very much for the feedback!

Hi @amilcarlucas,

It’s come up for quadplanes as well so I think it can affect any size of vehicle.

3 Likes
5

Big copters cost lot of money, nobody whant unexpected crash. But some later we will check new SW on big copter. After I’ll check same SW on my 15”

1 Like
6

I’m seeing something weird in acro with expo, I have sent @Leonardthall a log.

1 Like
7

I would expect it to be more noticeable on big vehicles.

EDIT: Really I mean vehicles with higher values of ATC_INPUT_TC which do tend to be the larger ones (but not always).

2 Likes
8

It looked like the ATC_ACC_P_MAX, ATC_ACC_R_MAX and ATC_ACC_Y_MAX all were reset to default.

The original were apparently:

ATC_ACCEL_P_MAX,681398.9

ATC_ACCEL_R_MAX,848569

ATC_ACCEL_Y_MAX,278635.7

9

@Leonardthall, @andyp1per,

I think this can happen if an early version of 4.7 was run on the vehicle before the AC_AttitudeControl::convert_parameters function was completed for 4.7. Put another way, there was a one month delay between when the ANGLE_MAX parameter was renamed/rescaled and when the ATC_ACC_xxx parameters were renamed/rescaled.

I think we could resolve this issue by removing these lines from AC_AttitudeControl::convert_parameters()

    // return immediately if no conversion is needed
    if (_angle_max_deg.configured()) {
        return;
    }

.. but to be clear, I think this only affects a small number of boards which had 4.7.0-dev installed about a month before the beta was released.

3 Likes
10

We use default 0,15 TC even on heavy drones. Just decreased att Rates and accelerations. On FPV I set TC 0,05.

2 Likes
11

Don’t do it just for me. The timeline sounds right - I upgraded two drones and this one was quite a bit earlier and I don’t think the other one had this problem - but I will check.

1 Like
12

Besides the fact that the drone did feel a bit “looser” on reaching a setpoint with my 4.6.3 position controller settings but is now resolved for me atleast by making the PSC P and I terms to default and TC to Crisp (I don’t remember if the latter change was necessary), the other way I gauged the loiter performance was by yawing.
I had sometime earlier posted an issue with my drone making translating movements while yawing (not yawing in place). So I almost solved it by tweaked PSC_I in 4.6.3 to 0.85 to get it to yaw in place. But when I upgraded to 4.7, the yaw behaviour has changed again so that the translation motion was present so I found that setting PSC_I to 1.15 solved it, even better than before, fortunately. This was all with yaw TC set to Medium and the roll/pitch TC to 0.15. Another point to note is that my optical flow is offset somthing like 7cm to 8.5cm behind the CoM so I had the FLOW_POS set in 4.6.3 to about roughly this much although not exact.

Afterwards I tested setting TC to 0.1 and the thing is a bit off again but down to the 4.6.3 levels and I find that setting FLOW_POS to 0 is better than the -7cm offset

1 Like